Surface chemistry and diffusion of trace and alloying elements during in vacuum thermal deoxidation of stainless steel

Removal of the native surface oxide from steel is an important initial step during vacuum brazing. Trace and alloying elements in steel, such as Mn, Si, and Ni, can diffuse to the surface and influence the deoxidation process. The detailed surface chemical composition and grain morphology of the common stainless‐steel grade 316L is imaged and spectroscopically analyzed at several stages of in‐vacuum annealing from room temperature up to 850°C. Measurements are performed using synchrotron‐based X‐ray photoemission and low‐energy electron microscopy (XPEEM/LEEM). The initial native Cr surface oxide is amorphous and unaffected by the underlying Fe grain morphology. After annealing to ~700°C, the grain morphology is seen at the surface, persisting also after the complete oxygen removal at 850°C. The surface concentration of first Mn and then Si increases significantly when annealing to 500°C and 700°C, respectively, while Ni and Cr concentrations do not change. Mn and Si are not located only in grain boundaries or clusters but are distributed across over the surface. Both Mn and Si appear as oxides, while Cr oxide becomes metallic Cr. Annealing from 500°C up to 850°C leads to the removal of first the Mn and then Si oxides from the surface, while Cr and Fe are completely reduced to metals. Deoxidation of Cr occurs faster at the grain boundaries, and the final Cr metal surface content varies between the grains. The findings are summarized in a general qualitative model, relevant for austenite steels.

Removal of the native surface oxide from steel is an important initial step during vacuum brazing. Trace and alloying elements in steel, such as Mn, Si, and Ni, can diffuse to the surface and influence the deoxidation process. The detailed surface chemical composition and grain morphology of the common stainless-steel grade 316L is imaged and spectroscopically analyzed at several stages of in-vacuum annealing from room temperature up to 850 C. Measurements are performed using synchrotronbased X-ray photoemission and low-energy electron microscopy (XPEEM/LEEM).
The initial native Cr surface oxide is amorphous and unaffected by the underlying Fe grain morphology. After annealing to $700 C, the grain morphology is seen at the surface, persisting also after the complete oxygen removal at 850 C. The surface concentration of first Mn and then Si increases significantly when annealing to 500 C and 700 C, respectively, while Ni and Cr concentrations do not change. Mn and Si are not located only in grain boundaries or clusters but are distributed across over the surface. Both Mn and Si appear as oxides, while Cr oxide becomes metallic Cr. Annealing from 500 C up to 850 C leads to the removal of first the Mn and then Si oxides from the surface, while Cr and Fe are completely reduced to metals.
Deoxidation of Cr occurs faster at the grain boundaries, and the final Cr metal surface content varies between the grains. The findings are summarized in a general qualitative model, relevant for austenite steels.

| INTRODUCTION
Stainless steels were introduced at the beginning of the 20th century, and in the ensuing half century, academia and manufacturers developed a large family of stainless steels serving many different applications. The austenitic (face centered cubic, FCC) stainless-steel grade 316L (UNS-S31603/EN-1.4404), studied in this work, is one of the most common iron-based alloys that is used for a wide range of applications from cookware to industrial equipment. On all stainless steels, a native oxide layer, usually termed "passive layer," acts as an excellent barrier against corrosion. It is generally accepted that the few nanometers thick protective oxide is a result of oxidation of the Cr (present in the base alloy) at the steel surface. 1,2 Beyond Cr (16-18% wt.), 316L is also alloyed with Ni (10-14% wt.), Mo (2-3% wt.), and Mn (<2% wt.). The primary purpose of Ni and Mn is to stabilize the austenitic structure while Mo increases the resistance to pitting corrosion. 2 In addition, low levels (<1% wt.) of several other elements are present in the steel such as Si (<1% wt.), P (<0.03% wt.), C (<0.03% wt.), and S (<0.045% wt.), which are trace elements from the miningand manufacturing process.
Many products manufactured in 316L require metal joining such as welding or brazing operations, which represent a significant global value 3 and are still highly active areas of research. 4 In the widely used high-volume manufacturing process, vacuum brazing, the steel is exposed to elevated temperatures in batch furnaces with a filler metal, usually consisting of copper or nickel alloys. During elevated temperature, the filler metal melts and wets the stainless steel forming a brazed joint. The molten filler metals are considered as high surface energy liquids. Wetting of the solid stainless steel with the molten filler metal is possible only if the interfacial bond between the two is strong, which requires a metallic or chemical character of the bond. Therefore, the presence of an oxide can influence the wetting during the brazing, resulting in a poor quality of the brazed joint, and failure can occur in the end product. Hence, a complete and controlled deoxidation of the native oxide is decisive for successful stainless-steel vacuum brazing operations. While oxidation of stainless steel has been investigated in significant detail, the deoxidation of the native oxide is much less studied.
The temperature-dependent stability of compounds of metals, its oxides, and oxygen can generally be found in the Ellingham diagram. 5,6 However, considering a modern full-scale industrial vacuum brazing furnace, the O partial pressure level will never be enough to permit reduction of the chromium oxide according to the Ellingham diagram. Instead, there exist several explanations how the initial deoxidation transpires such as difference in thermal expansion 6 and surface diffusion of carbon. 7 However, because of the challenges in characterizing a nanometer thin oxide in combination with high temperature, no consensus in this area exists. A detailed understanding of the deoxidation and the role of both the alloying and trace elements are essential for a more efficient and controlled manufacturing, and the development of new metal joining methods and stainless steels optimized for vacuum brazing.
In the last 20 years, X-ray photoelectron spectroscopy and photoemission electron microscopy (XPS and XPEEM) were used in several studies of the surface of stainless steel. [8][9][10][11][12][13][14][15][16][17] XPS was used to reveal the spatially averaged surface elemental and sometimes chemical composition. [8][9][10][11][12] A number of these studies focused on the oxidation of duplex steels, [8][9][10] where XPS could reveal the composition and oxidation states of the surface's oxides after various stages of controlled oxidation. The surface segregation during annealing of the 316L steel was studied using laboratory based XPS 11 ; however, in this study, the surface oxide was removed by sputtering prior to annealing. Although sputtered surface allows a clear interpretation of the XPS data, it makes it difficult to directly relate to the diffusion conditions in the presence of the native Cr surface oxides. With the advent of XPEEM, it has become possible to combine the spectroscopic information from photoemission with lateral imaging down to tens of nanometers. [13][14][15][16][17] A particular advantage of XPEEM is that the (potentially significant) variations in the chemical composition due to the grain structure of the metal and presence of inclusions become directly visible. Some XPEEM studies show the elemental and oxide distribution across the metal surfaces. [13][14][15][16] A considerable focus in previous XPEEM work has been on oxidation under various controlled O 2 pressures. 16 Additionally, a hard X-ray PEEM (HAXPEEM) study that extends the depth profiling on duplex steel using PEEM has been carried out. 14 Finally, a PEEM study of the 316L steel, 15 in which the sample was sputtered and annealed prior to imaging, revealed interesting information on the steel grains, but it did not provide information on the surface deoxidation.
In the present study, we focus on the deoxidation process, when the steel samples (with the native Cr-oxide present) are heated in vacuum. XPEEM provides information on the depth and lateral elemental as well as chemical state distribution at the surface, during the native oxide removal. Low-energy electron microscopy (LEEM) provides information on the morphology of the surface, visualizing individual grains. We prepare samples by specialized polishing to allow the sample to be imaged with high voltage in XPEEM and LEEM. No additional treatment is performed prior to measurements; thus, the samples have a native oxide close to the one found under manufacturing conditions. XPEEM allow XPS and X-ray absorption spectroscopy (XAS) of trace and alloying elements with extremely high surface sensitivity and nanoscale spatial resolution. As the sample can be heated in vacuum, we can follow the deoxidation with XPEEM/LEEM, observing how Mn and Si (which are present in small amounts in the bulk) diffuse to the surface and form oxides using the oxygen from the native Cr oxides. Vacuum annealing is relevant to industrial process as it is also used during vacuum brazing. In the present case, the vacuum pressure is lower than under industrial conditions. However, as it has previously been found that the qualitative behavior of oxide stability in the vacuum regime is similar across several orders of magnitude in pressure, while the precise transition temperature in the phase diagram might change. 18,19 Although the vacuum level pressure of the present study is lower than in the industrial vacuum brazing process, it still provides qualitative information on the dynamic process as discussed below.

| Spectromicroscopy
Photoemission and low-energy electron microscopy experiments were performed at MAXPEEM, a dedicated beamline at MAX IV Laboratory. The beamline houses a state-of-the-art aberration-corrected spectroscopic photoemission and low-energy electron microscope (SPELEEM). This powerful instrument offers a wide range of complementary techniques providing structural, chemical, and magnetic sensitivity with a single-digit nanometer spatial resolution. This includes XPEEM and LEEM as well as XPS and XAS spectroscopy in selective areas down to a few tens of square nanometers. The beamline can deliver a high photon flux in the range 30-1200 eV. In the MAXPEEM setup, the X-ray beam impinges on the sample at nominal incidence.
The microscope is equipped with an aberration corrector which improves both the resolution and transmission. As a (photo)electron detector, a new CMOS TVIPS-F216 camera is used. In the present work, two different operation modes of the microscope have been used: microscopy and spectroscopy modes. In the microscopy mode, the image plane of the objective lens is displayed on the screen. The energy slit is placed at the dispersive plane of the energy analyzer, selecting only electrons with a given kinetic energy. The kinetic energy of the photoelectrons is set by applying a voltage bias, referred to as "start voltage," to the sample. In the spectroscopy mode, the projector magnifies the dispersive plane at the end of the energy filter onto the screen. The intensity line profile over the spread (16 eV) electron beam reveals the energy distribution of the photoelectrons.
Annealing was performed using electron bombardment filament placed behind the sample. Temperature was measured with a thermocouple placed just below the heater and an infrared pyrometer via a viewport. The base pressure in the measurement chamber is 1 Â 10 À10 mbar, during annealing the pressure raised to a maximum of 5 Â 10 À9 mbar. Low temperature (≤550 C) annealing of the samples was carried out in the preparation chamber for 30 min. Hightemperature (>550 C) annealing was carried out in the measurement chamber for a duration of 5 min in order to maintain the pressure in the low 10 À9 mbar range. Subsequently, all spectra were recorded at room temperature.

| Sample preparation
Bulk metal was cut to cuboids sized 8 mm Â 8 mm Â 1 mm using a

Struers Secotom 50 precision cutter and embedded in Struers Black
MultiFast hot mounting resin (thermosetting bakelite with wood filler) at 180 C using Struers CitoPress-5 electro-hydraulic hot press. The samples were embedded with one face exposed on the bakelite surface. The exposed faces were polished using a Struers Tegramin-30 microprocessor controlled grinder set to pad rotation at 150 rpm and holder co-rotating at 150 rpm using the following recipe: 3 | RESULTS AND DISCUSSION

| Initial visualization of the amorphous oxide removal and appearance of grains
To get an initial overview of surface topography of the 316L steel during annealing, we use LEEM in the so-called Mirror mode (Figure 1). In Mirror mode, the sample bias voltage is adjusted to allow minimal interaction of the electron beam with the surface. The contrast is determined by local variation in work function or height across the sample surface. Thus, any changes in morphology or local atomic scale structural termination (that leads to work function changes) of the surface will be visible.
In Figure 1, the surface of the 316L sample is imaged live upon annealing. Up to $620 C, the features in the images consist of a few lines and dark spots which we attribute to scratches and minor debris from the polishing (consistent with AFM images of steel samples after polishing, see Figure S1). As the process of oxide transformation/ removal starts to occur, the underlying bulk grain structure becomes visible, as seen in Figure 1C,D. As this imaging mode is highly surface sensitive, we interpret this as an amorphous oxide being present initially, which is unaffected by the underlying grain structure and homogeneous on the resolution level of the LEEM images. As this oxide is reduced, the underlying grain structure, including boundaries, manifests in LEEM image due to diffraction contrast. The austenite FCC crystal grains exhibit different intensity depending on surface crystalline structure and orientations. In summary, the Cr oxide, which is covering the metal surface at low temperatures, is a continuous amorphous film lacking both grain boundaries and surface crystalline structure. After high temperature deoxidation, the grain boundaries of the underlying austenite Fe become visible. It can further be noted that up to 850 C, the grain morphology itself appears stable. To further elucidate the chemical changes of the surface oxide and the relation to the grains, we will turn to spectroscopic imaging in the next section.

| Elemental and chemical surface distribution as a function of annealing temperature
The surface chemistry was studied after annealing at several temperatures up to 850 C by chemical analysis of trace and alloying elements.
Chemical analysis was performed in two different spectroscopic modes: direct photoemission (XPS) and X-ray absorption spectroscopy (XAS). In the XPS mode, the microscope is tuned to record intensity of an elemental core-level of interest. In the XAS mode, the photon energy is scanned over the elemental absorption edge while the change in secondary electron emission is recorded. The direct XPS mode is more surface sensitive typically probing 1-2 nm, while the absorption mode has a probing depth of 5-10 nm. 20 In Figure 2, we show the average spectroscopic signal across the surface of four elements as a function of annealing temperature. At room temperature, the main signal at the surface comes only from Cr in the form of metallic Cr and Cr oxide (a double peak at 577 eV in the Cr L-edge XAS spectrum), as shown in Figure 2C. For reference, we briefly mention overview XPS measurements done in previous studies of the mechanically polished 316L alloy prior to heat treatments. 12,22,23 Compared with the bulk ratios for this alloy, the Cr/Fe ratio is $3 times larger at the surface compared with the bulk composition, while the relative amount of Ni is similar to the bulk values.
Other elements that exist in small amounts inside the alloy can also be detected in small amounts at the surface. Returning to the present study, we find that upon annealing, the Cr oxide component markedly decreases and disappears above 850 C. The oxide evolution can also be traced by recording the direct photoemission signal from O. Figure 2A shows XPS measurements of O1s peak as a function of temperature. Upon annealing, the O1s peak area is steady up to 570 C, at which point it decreases with $50% from 570 C to 700 C and then the peak disappears at 850 C (indicating complete surface oxide removal). The peak position shifts towards higher binding energy as the temperature increase. In particular, the significant change at 700 C coincides with the appearance of a strong Si2p peak appearing (see Figure 2B). A steady shift of the O1s peak towards higher binding energies together with a similar shift of the with the direct evaporation of the Si-oxide. 25 Mn has a high vapor pressure at these temperatures, which would be in favor of the element evaporating. 11,26,27 F I G U R E 1 (A-D) MEM images of the same surface annealed at different temperatures (120 C, 620 C, 700 C, and 850 C, respectively). The annealing temperature is shown at the upright corner of image. The grain boundaries start to be seen in (C), indicating that it occurred between 620 C and 700 C. LEEM (electron kinetic energy: 3.15 eV) image obtained after annealing at 850 C. The contrast coming from the crystal structure of the grains at the surface, which is different due to their different orientations. Images on (A-D) are of identical size, the scale bars as seen in (A) and (E) are 10 μm Previous XPS measurements on stainless steels also found surface out-diffusion of Mn and Si during heating. 11 In that study, sputtering was conducted prior to annealing resulting in a very different initial surface composition. This could indicate that the elemental out-diffusion of Mn and Si is a general phenomenon independent of initial surface conditions. Further, no oxygen was initially present in, 11 but some O emerged at the surface at later stages of the annealing.
Thus, the native oxides could also play a role in stabilizing the metal impurities at the surface. A significant increase in Cr content found in Gröning et al. 11 is not observed in the present study, presumably due to their initial removal of the Cr-oxide at the surface. The different results indicate that a preferred elemental and chemical balance is likely to exist at the surface related to energetics.
Cr-oxide decomposition can be followed in more detail by twocomponent fitting of the Cr L-edge absorption spectra (Figure 3). The spectra at lowest (120 C) and highest (850 C) temperatures were used as reference for oxide and metallic Cr, assuming that Cr is 100% and only metal Cr is left, which is consistent with no peak being observed in the O1s spectra. Finally, Ni and Fe spectroscopic imaging was also performed, and average spectra deduced as seen in Figure 4.
The shape of the Fe L-edge after annealing at 120 C (Figure 4, left, blue solid curve) can be associated with the presence of oxidized Fe, 28 while the spectra observed after annealing to 750 C (Figure 4, left, orange dash curve) is more similar to metallic Fe. 28 The missing O can be either taken up by the Mn and Si (as discussed above) or evaporate into the vacuum. The amount of Ni at the surface appears to decrease as the sample is annealed, and no significant chemical change is observed for the Ni. Here, it can be noted that previous ab initio calculations of alloy compounds have found that while Ni segregation in metallic alloys can be favorable at room temperature, this trend starts to reverse at higher temperatures. 29 Another similar study 30 has found that the presence of O strongly influences the segregation of Cr at room temperature, leading to the segregation and high abundance of Cr.
F I G U R E 2 Core-level X-ray photoemission spectra (XPS) for O1s (A) and Si2p (B) acquired after annealing at different temperatures. X-ray absorption spectra (XAS) for Cr L-edge (C) and Mn L-edge (D). Cr L-edge spectra (C) upon annealing display a smooth transition from Cr oxide (the blue curve, 120 C) to a metallic Cr (the red curve, 850 C). Mn L-edge XAS show manganese segregation at the surface in the form of manganese oxide. The signal drops markedly, and no traces of manganese can be detected at 850 C. In the MnO XAS plot, a reference spectrum for Mn oxide (black dash line) is presented with a small (0.6 eV) energy shift 21

| Lateral distribution of elements and their chemical states at surface
An important question that can be answered using XPEEM is the distribution of elements and chemical species laterally across the surface.
Prior to the grains becoming visible at 650 C-700 C, the featureless appearance of the MEM and XPEEM images indicates that the chemical components are homogeneously distributed across the surface (except differences due to the polishing scratches) on the length scales visible in the PEEM images (20 μm-10 nm). Figure 5A shows a F I G U R E 3 XAS spectra fitting of Cr based on the XAS spectrum for Cr L-edge with background removed at different temperatures as seen in Figure 2C. The spectra at lowest (120 C) and highest (850 C) temperatures are taken as references for oxide and metallic Cr, respectively. The spectra fitting (linear regression) with two components (the oxide and metal ones) are shown for 550 C (A) and 700 C (B). The fitting coefficients are shown in the legend. (C) The areas of Cr 2 O 3 (blue shade) and metallic Cr (red shade) were used to calculate the percentage of Cr 2 O 3 and metallic Cr at different temperatures F I G U R E 4 XAS for Fe L-edge (left) and Ni L-edge (right) from sample AISI 316L after annealing at 120 C and 750 C. (left) The Fe L-edge exhibits sharp 2p3/2 and 2p1/2 peaks where shoulders at higher photon energies can be seen after annealing at 120 C. These extra peaks can be associated with the presence of an iron oxide 28 secondary electron image recorded using electrons with a photon energy of 577 eV (the Cr oxide maximum in the XAS spectra) after annealing to 700 C. The grain boundaries can be clearly seen. However, to robustly analyze the difference between the grain boundary and grain, we extract spectra from a grain center and a grain boundary, respectively. The results are shown in Figure 5B where the intensity is plotted as a function of photon energy. The areas used to generate the spectra were of the same size and close to each other, so that they can be properly compared. These results show that after annealing at high temperatures, a larger content of metallic Cr is present along grain boundaries. Further insights can be gained from measuring XAS spectra of Cr from several grains after annealing to 850 C (after complete removal of O), as seen in Figure 5C,D. It can be seen that the Cr content in the surface region of the grains varies significantly from grain to grain. A below threshold image ( Figure 5C) also shows a grain boundary contrast, which is consistent with the fact that the work function (WF) of the metallic Cr (4.5 eV) is lower than the WF of Iron (4.7-4.8 eV). 27 So after annealing at high temperatures, we observe segregation of metallic Cr along grain boundaries.
For Si and Mn, a more qualitative analysis was done in the temperature range between 500 C and 700 C when they are abundant on the surface. This is done by comparing images recorded at the core-level peak/adsorption edge of the materials versus off the peak as shown in Figures S3 and S4. At 550 C, before the grain boundaries

| A qualitative model of the deoxidation
We now summarize the findings in a qualitative model of the deoxidation process by in-vacuum annealing, as seen in Figure 6 show that SiO 2 is highly energetically favorable compared with Cr, Fe, and Mn oxides. In conclusion, the presented model and observed data are consistent with available theoretical calculations although for the complete understanding of the Mn and Si segregation as well as preferential oxidation further calculations would be relevant. Ab initio calculations appear to form a good basis for understanding these systems, but it would be interesting to investigate when for example kinetic effects will start to play a role, as has been seen in the oxide formation in single element alloys. 19 While the crystalline Fe grains can be observed at the surface as the Cr oxide begin to decompose, this initial protective oxide and the initial distribution of trace elements is amorphous and homogeneous in distribution in the tens of microns to tens of nanometers range.
Already, during annealing, there is a tendency of the Cr to gather in the grain boundaries of the Fe, with these boundaries being particularly rich in metallic Cr. After annealing, there is a nonhomogeneous Cr distribution at the surface among different grains and with some Cr segregation along the grain boundaries. In contrast, Si and Mn appear homogeneously distributed over the surface before they are removed at high temperatures.

| PERSPECTIVES
Elucidating the stainless-steel surface decomposition and role of alloying and trace elements can potentially be used to optimize the vacuum furnace programs, generating shorter lead times and more efficient manufacturing. However, while oxidation has been studied more extensively, the important initial step of deoxidation has been much less studied. In particular, it is important to begin with the initial native oxide that is also the starting point when processing the steel.
The study indicates that the deoxidation can potentially be significantly altered with tailored addition of small amounts of elements of both the stainless steel. Future studies of different steel compounds in the context of the deoxidation would be interesting to further investigate the influence of trace compounds, as they will affect each other as they rapidly diffuse across and into the surface while exchanging for example O.